AI For SMEs with Kauz.ai
AI for SMEs: Building Reliable, Production-Ready Agentic AI
Join us for AI Düsseldorf’s final summer session, hosted by Kauz.ai in Düsseldorf-Bilk.
As companies move from AI experimentation toward real-world implementation, the challenge is no longer simply accessing powerful language models. Organizations must also determine how to test AI systems, design reliable agent architectures and manage the growing trade-offs between model quality, performance and cost.
This session brings together three complementary perspectives on building agentic AI systems for production: evaluation and safety, architecture patterns, and intelligent model routing.
Speakers and talks
Harry Cruz
Founder & CTO, RhesisAI
Conversational AI for SMEs: Testing & Evaluation
Conversational AI is moving into production, but many organizations still lack a practical way to determine whether these systems are reliable, robust and safe.
Harry will present a practical approach to testing and evaluating conversational AI before deployment, including:
reliability and robustness testing
risks represented in the OWASP LLM threat model
evaluation tools for teams with limited resources
realistic steps SMEs can implement before going live
Dr. Ali Zendegani
Senior Data Scientist, Miele
Building Agentic AI: Architecture Patterns from the Claws Ecosystem
Drawing on a comparative analysis of 15 open-source “Claw” runtimes, Ali will highlight the architecture patterns that repeatedly appear across modern agentic AI systems.
The talk will explore:
execution loops and how agents plan and act
memory architectures and retrieval strategies
tool integration and orchestration
governance, safety and auditability
The session will provide a practical architectural perspective for companies moving beyond isolated prototypes toward reliable agentic systems.
Dr. Thomas Rüdel
Founder & CEO, Kauz.ai
Intelligent LLM Routing: Optimizing Quality and Cost in Agentic AI Systems
As agentic AI drives token usage and operational costs, using the same model for every task is rarely the most efficient approach.
Thomas will show how Kauz.ai uses its own evaluations to support intelligent model-selection decisions and balance:
model quality and task complexity
token usage and operational cost
dynamic LLM selection and routing
evaluation-driven trade-offs in production
This talk will offer a practical perspective on building AI systems that select the appropriate model for each task rather than defaulting to the most powerful—and expensive—option.
Agenda
18:30 – 18:45
Arrival and networking
18:45 – 18:55
Welcome and introduction by AI Düsseldorf and Kauz.ai
18:55 – 19:25
Harry Cruz — Conversational AI for SMEs: Testing & Evaluation
Including Q&A
19:25 – 19:55
Dr. Ali Zendegani — Building Agentic AI: Architecture Patterns from the Claws Ecosystem
Including Q&A
19:55 – 20:10
Break and networking
20:10 – 20:40
Dr. Thomas Rüdel — Intelligent LLM Routing: Optimizing Quality and Cost in Agentic AI Systems
Including Q&A
20:40 – 21:00
Closing discussion and networking
📅 30 July 2026
🕡 18:30–21:00
📍 Kauz.ai, Erasmusstraße 15, 40223 Düsseldorf
🏢 Hosted by Kauz.ai
The session is particularly relevant for AI and data leaders, technical decision-makers, product teams and practitioners working on reliable, cost-conscious and production-ready AI systems.